23
23
# TODO: Perhaps return more details on the file to avoid statting it
24
24
# again: nonexistent, file type, size, etc
26
# TODO: Perhaps use a Python pickle instead of a text file; might be faster.
30
28
CACHE_HEADER = "### bzr hashcache v5\n"
32
30
import os, stat, time
34
from bzrlib.osutils import sha_file, sha_string, pathjoin, safe_unicode
32
from bzrlib.osutils import sha_file
35
33
from bzrlib.trace import mutter, warning
36
from bzrlib.atomicfile import AtomicFile
37
from bzrlib.errors import BzrError
37
def _fingerprint(abspath):
39
fs = os.lstat(abspath)
41
# might be missing, etc
44
if stat.S_ISDIR(fs.st_mode):
47
# we discard any high precision because it's not reliable; perhaps we
48
# could do better on some systems?
49
return (fs.st_size, long(fs.st_mtime),
50
long(fs.st_ctime), fs.st_ino, fs.st_dev)
46
53
class HashCache(object):
104
111
self.needs_write = True
108
"""Scan all files and remove entries where the cache entry is obsolete.
110
Obsolete entries are those where the file has been modified or deleted
111
since the entry was inserted.
113
# FIXME optimisation opportunity, on linux [and check other oses]:
114
# rather than iteritems order, stat in inode order.
115
prep = [(ce[1][3], path, ce) for (path, ce) in self._cache.iteritems()]
115
def refresh_all(self):
116
prep = [(ce[1][3], path) for (path, ce) in self._cache.iteritems()]
118
for inum, path, cache_entry in prep:
119
abspath = pathjoin(self.root, path)
120
fp = self._fingerprint(abspath)
123
cache_fp = cache_entry[1]
125
if (not fp) or (cache_fp != fp):
126
# not here or not a regular file anymore
127
self.removed_count += 1
128
self.needs_write = True
129
del self._cache[path]
131
def get_sha1(self, path, stat_value=None):
119
for inum, path in prep:
120
# we don't really need to re-hash them; we just need to check
121
# if they're up to date
125
def get_sha1(self, path):
132
126
"""Return the sha1 of a file.
134
if path.__class__ is str:
135
abspath = pathjoin(self.root_utf8, path)
128
abspath = os.sep.join([self.basedir, path])
129
fp = _fingerprint(abspath)
131
c = self._cache.get(path)
133
cache_sha1, cache_fp = c
137
abspath = pathjoin(self.root, path)
135
cache_sha1, cache_fp = None, None
138
137
self.stat_count += 1
139
file_fp = self._fingerprint(abspath, stat_value)
142
# not a regular file or not existing
143
141
if path in self._cache:
144
142
self.removed_count += 1
145
143
self.needs_write = True
146
144
del self._cache[path]
149
if path in self._cache:
150
cache_sha1, cache_fp = self._cache[path]
152
cache_sha1, cache_fp = None, None
154
if cache_fp == file_fp:
155
## mutter("hashcache hit for %s %r -> %s", path, file_fp, cache_sha1)
156
## mutter("now = %s", time.time())
146
elif cache_fp and (cache_fp == fp):
157
147
self.hit_count += 1
158
148
return cache_sha1
162
mode = file_fp[FP_MODE_COLUMN]
163
if stat.S_ISREG(mode):
164
digest = self._really_sha1_file(abspath)
165
elif stat.S_ISLNK(mode):
166
digest = sha_string(os.readlink(abspath))
168
raise BzrError("file %r: unknown file stat mode: %o"%(abspath,mode))
151
digest = sha_file(file(abspath, 'rb', buffering=65000))
170
# window of 3 seconds to allow for 2s resolution on windows,
171
# unsynchronized file servers, etc.
172
cutoff = self._cutoff_time()
173
if file_fp[FP_MTIME_COLUMN] >= cutoff \
174
or file_fp[FP_CTIME_COLUMN] >= cutoff:
175
# changed too recently; can't be cached. we can
176
# return the result and it could possibly be cached
179
# the point is that we only want to cache when we are sure that any
180
# subsequent modifications of the file can be detected. If a
181
# modification neither changes the inode, the device, the size, nor
182
# the mode, then we can only distinguish it by time; therefore we
183
# need to let sufficient time elapse before we may cache this entry
184
# again. If we didn't do this, then, for example, a very quick 1
185
# byte replacement in the file might go undetected.
186
## mutter('%r modified too recently; not caching', path)
187
self.danger_count += 1
189
self.removed_count += 1
153
now = int(time.time())
154
if fp[1] >= now or fp[2] >= now:
155
# changed too recently; can't be cached. we can
156
# return the result and it could possibly be cached
158
self.danger_count += 1
160
self.removed_count += 1
161
self.needs_write = True
162
del self._cache[path]
163
elif (fp != cache_fp) or (digest != cache_sha1):
164
# mutter("update entry for %s" % path)
165
# mutter(" %r" % (fp,))
166
# mutter(" %r" % (cache_fp,))
190
167
self.needs_write = True
191
del self._cache[path]
193
## mutter('%r added to cache: now=%f, mtime=%d, ctime=%d',
194
## path, time.time(), file_fp[FP_MTIME_COLUMN],
195
## file_fp[FP_CTIME_COLUMN])
196
self.update_count += 1
197
self.needs_write = True
198
self._cache[path] = (digest, file_fp)
201
def _really_sha1_file(self, abspath):
202
"""Calculate the SHA1 of a file by reading the full text"""
203
return sha_file(file(abspath, 'rb', buffering=65000))
168
self._cache[path] = (digest, fp)
206
179
"""Write contents of cache to file."""
207
outf = AtomicFile(self.cache_file_name(), 'wb', new_mode=self._mode)
180
from atomicfile import AtomicFile
182
outf = AtomicFile(self.cache_file_name(), 'wb')
209
outf.write(CACHE_HEADER)
184
print >>outf, CACHE_HEADER,
211
186
for path, c in self._cache.iteritems():
212
line_info = [path.encode('utf-8'), '// ', c[0], ' ']
213
line_info.append(' '.join([str(fld) for fld in c[1]]))
214
line_info.append('\n')
215
outf.write(''.join(line_info))
187
assert '//' not in path, path
188
outf.write(path.encode('utf-8'))
190
print >>outf, c[0], # hex sha1
192
print >>outf, "%d" % fld,
217
196
self.needs_write = False
218
## mutter("write hash cache: %s hits=%d misses=%d stat=%d recent=%d updates=%d",
219
## self.cache_file_name(), self.hit_count, self.miss_count,
221
## self.danger_count, self.update_count)
226
204
"""Reinstate cache from file.
added
removed
bzrlib/hashcache.py
